Cloudflare said that during Birthday Week it made Logpush available on all self-serve plans and made Transformers generally available1. As the company describes it, a Transformer is a SQL query written for a specific Logpush dataset. Records that fail the WHERE clause are dropped. Remaining records can have fields redacted, renamed, added, removed or nested before export as NDJSON1. Cloudflare says this removes the need for a separate processing service and can reduce downstream data volume and cost1. Queries run inside the delivery service on the Apache DataFusion engine. Failures are handled "fail closed": untransformed data is not sent, retries are progressively delayed and persistently failing jobs are automatically disabled1. The cited performance figure, about 3 ms for a synthetic 5,000-row batch, comes from the company's own benchmark1. In the editors' assessment, actual savings will depend on log volume and composition. To try it, Cloudflare suggests opening Transformer Studio, choosing a dataset, previewing the query on a sample record, saving it and attaching it to a matching job. The feature is offered on Free, Pro, Business and Enterprise plans1.

  1. [1] Filter, reshape, and redact logs before export with TransformersCloudflare